Many companies are struggling to prepare their organizations for generative AI. As they navigate this process, they typically choose from one of three models: the centralized, decentralized or open approach. But which one is best?
Our data reveals that leading companies across industries aren’t struggling. They’ve embraced innovation and embedded it into their everyday operations. GenAI is revolutionary for these outperformers and just another tool to integrate seamlessly into their workflow.
What sets these companies apart is their customer-centric approach to genAI. Instead of organizing around a specific model, they blend all three, knowing when to prioritize each.
This flexible strategy allows them to avoid the common pitfall of introducing new technology without a clear purpose. Instead, they focus on delivering the right solution to meet customer needs.
3 frequently used genAI-integration models
Below are three distinct models companies use to integrate genAI capabilities into their organizations:
Open model
This is the most flexible approach, where genAI tools are available to anyone in the organization with minimal oversight. It encourages rapid innovation and adoption but also poses compliance and governance risks. The open model works best when experimentation is encouraged within set boundaries, relying on trust and the guideline: “Don’t do stupid things.”
Decentralized model (Labs)
The decentralized model allows different departments to experiment with genAI independently. This model is sometimes referred to in organizations as “Labs.” It fosters the agility that specialist teams need to quickly test and iterate on new ideas without waiting for approval from a central authority. However, if design principles are not adhered to, it can lead to fragmentation and inconsistencies in AI deployments.
Centralized model
In this approach, genAI initiatives are managed by one dedicated team, often within IT or a dedicated AI department. This allows for consistent governance, streamlined processes and a unified strategy. However, distributing concepts into the organization can also lead to bottlenecks. The centralized model is ideal for organizations that require strict control over AI deployments, such as those in highly regulated industries.
Dig deeper: Integrating AI into MOps: Aligning your platforms, data and processes
Why blend the three models?
The outperformers use these three models in tandem because they understand that they serve the company and the customer differently. Each has different strengths and weaknesses that should be considered.
| Model | Good for | Bad for | Need for |
| Open | Speed | Compliance | IP and legal guardrails |
| Decentralized | Relevance | Fragmentation | Design principles |
| Centralized | Control | Proliferation | Predictability and scalability |
From a customer’s perspective, proposition maturity is key. Newer propositions benefit from an open model to encourage innovation while still using a centralized model for legal compliance. More mature propositions with proven business cases need a central model to ensure scalability and predictability.
The open model encourages experimentation, while the centralized model focuses on exploitation. The latter offers standards and guidelines, especially for design principles, intellectual property and legal guardrails.
The decentralized model acts as a bridge between the two. The open model fosters innovation, but its fragmented nature can prevent that innovation from fully developing. The decentralized model helps ideas mature before integrating them into production using the centralized model.
Hack, pack and/or stack?
The outperformers learned that each model has a different goal, methodology, support, development and mindset. To emphasize that the models work in tandem, let’s call these three stages “hack, pack and stack.”
| Model | Goal | Process approach | Methodology approach | IT approach |
| Hack | Problem-Market fit | Project | Design Thinking | PoC / Prototype |
| Pack | Product-Market fit | Process | Lean Startup | MVP |
| Stack | Platform-Market fit | Product | Agile (scrum) | Production |
Hack: The art of experimentation
Hacking involves rapid experimentation through one-off projects, much like the campaigns we’re used to. It focuses on creating standalone versions, proof of concepts (PoCs) and prototypes to test technical feasibility, data viability and customer interest.
By applying design thinking, you can identify and iterate on the unique customer experience — those critical moments that differentiate your company. The goal isn’t to find a perfect solution, but to achieve a problem-market fit that resonates with customers, demonstrates traction and presents a solid business case. This is where startups excel.
Once you’ve established the business case, you’re ready to move on to the next phase: packing.
Pack: The art of scaling
With the customer problem clearly defined, the next step is to explore relevant genAI solutions to achieve product-market fit. The hack version undergoes packaging, resulting in a standardized process around the product. This involves refining the initial version by eliminating redundant features and data points and applying established IT design principles.
A highly effective approach at this stage is to build minimum viable products (MVPs) to test core functionalities. This enables teams to make necessary adjustments before moving on to full-scale development.
Stack: The art of exploitation
With the customer product in focus, the aim is to achieve platform-market fit. By removing redundant data, features and integration points using company design rules, the packaged version becomes ready for integration into the production stack. The development follows an iterative approach, using Agile (Scrum) methodology to break work into sprints for continuous improvement and adaptability.
Once the core platform is validated, the focus shifts to scaling it for production, ensuring it is robust and ready for full deployment with minimal maintenance. This frees IT resources, prevents legacy issues and allows teams to focus on the next innovation experiment.
Dig deeper: The truth behind martech stack composability
The three steps
The “hack, pack and stack” mindset offers a dynamic framework for martech teams to innovate, scale and integrate genAI solutions effectively.
- The “Hack-version”
- Create a stand-alone version to find out if it can be done technically and data-wise and if the customer likes it.
- The “Pack-version”
- Once there is proper customer traction, clean up the hack by removing anything that can be removed (data, content, lists, ETL).
- The “Stack-version”
- Refactor into a scalable zero-maintenance version and integrate into the ecosystem.
Adopting this flexible approach will be crucial for staying competitive and delivering AI-enhanced customer experiences.
Dig deeper: Where to deploy AI for maximum martech impact
Contributing authors are invited to create content for MarTech and are chosen for their expertise and contribution to the martech community. Our contributors work under the oversight of the editorial staff and contributions are checked for quality and relevance to our readers. The opinions they express are their own.
